Publication number | US7769217 B2 |
Publication type | Grant |
Application number | US 11/714,406 |
Publication date | Aug 3, 2010 |
Filing date | Mar 6, 2007 |
Priority date | Mar 6, 2006 |
Fee status | Lapsed |
Also published as | US20070230762 |
Publication number | 11714406, 714406, US 7769217 B2, US 7769217B2, US-B2-7769217, US7769217 B2, US7769217B2 |
Inventors | James J. Hamill |
Original Assignee | Siemens Medical Solutions Usa, Inc. |
Export Citation | BiBTeX, EndNote, RefMan |
Patent Citations (3), Non-Patent Citations (4), Referenced by (1), Classifications (6), Legal Events (4) | |
External Links: USPTO, USPTO Assignment, Espacenet | |
This application claims priority under 35 U.S.C. §119(e) of copending provisional application Ser. No. 60/779,530 filed Mar. 6, 2006.
1. Field of the Invention
The present application relates to nuclear medical imaging, and more particularly to methods and apparatus for reconstructing nuclear medical images from three-dimensional PET acquisition data using linograms.
2. Related Art
Gamma cameras, also referred to as nuclear imaging cameras, radioisotope cameras, scintillation cameras and Anger cameras, are used to measure gamma radiation emitted by a body under examination. By measuring the energy and the location of the gamma ray emissions, an image representative of the gamma radiation emitted from the body under examination can be created.
In the field of positron emission tomography (PET), opposing detectors detect the emission from annihilation of a positron of a pair of simultaneous gamma photons traveling in opposite directions. Typically, a large number of scintillation detectors (e.g., scintillation crystal detectors) are arranged in rings. The PET electronics determine when two oppositely arranged detectors produce an event signal at essentially the same time, indicating that a pair of annihilation gamma rays has been detected. The emitted positron that was the source of the pair of gamma rays is considered to be located on a line of response (LOR) connecting the locations on each detector where each gamma ray was detected. The principle is illustrated by
The emission source causes coincidences between detectors A and D, and between B and C, but not between A and C, or B and D. The x and y coordinates of the source (see
In PET, one attempts to reconstruct three dimensional objects by using measurements of line integrals through the measurement space.
In practice, instrument designers increase the z sampling by operating the detectors three dimensionally. Coincidences are recorded between detectors that are not in a common plane. If one detector has coordinate z_{1 }and a coincident detector has coordinate z_{2}, the simplest approximate formula for the source z coordinate is z_{s}=(z_{1}=z_{2})/2. This approximation has been called single-slice rebinning. The well-known Fourier rebinning method or FORE, is another approach, and is still used in most of the clinical 3D PET performed around the world.
Grouping lines of response into a plane of constant z, with techniques like the ones mentioned in the last paragraph, leads to faster reconstructions. Two-dimensional reconstruction methods are normally much faster than 3D reconstructions. The speed advantage is important because current PET image reconstruction uses iterative techniques, in which image fidelity tends to improve as one uses more iterations, yet one is forced to stop the iterations before convergence is attained, so that the image can be delivered to the physician in an acceptable amount of time.
It is well-known that the ML-EM (Maximum Likelihood Expectation Maximization) (or simply “EM”) iterative algorithm provides very good PET image reconstruction, but it has been considered too slow for clinical use. A primary drawback for using the ML-EM method is that it requires, in principle, an infinite number of cycles, or iterations, to converge to the answer. The subsequent development of acceleration techniques such as Ordered Subset Expectation Maximization (OSEM), and the development of Fourier rebinning for transforming 3D PET sinograms into 2D sinograms, made iterative reconstruction clinically feasible by allowing approximations to be used in the algorithm. However, factors such as convergence and the limit-cycle effect still presented arguments in favor of EM.
The present inventors have previously provided a method combining EM with the Approximate Discrete Radon Transform (ADRT), as disclosed by M. L. Brady in “A fast discrete approximation algorithm for the Radon transform,” SIAM J. Comput., Vol. 27, no 1, p 107-119, February 1998. See Published U.S. Patent Application No. 2003/0190065, published Oct. 9, 2003, incorporated herein by reference in its entirety. This new method allowed EM reconstructions to be accelerated by orders of magnitude, thereby generating results in about the same time as OSEM.
However, there remains a need in the art to accelerate image reconstruction calculations even further, and to provide image reconstruction methods for 3D PET images.
The present invention improves upon the prior art by providing a PET image reconstruction method that is significantly faster than known methods such as OSEM, and further is capable of reconstructing 3D PET image data.
The preferred embodiments of the present invention are shown by a way of example, and not limitation, in the accompanying figures, in which:
While the present invention may be embodied in many different forms, a number of illustrative embodiments are described herein with the understanding that the present disclosure is to be considered as providing examples of the principles of the invention and that such examples are not intended to limit the invention to preferred embodiments described herein and/or illustrated herein.
The usual organization of 3D PET data can be explained in terms of the Orlov sphere.
Whereas 3D PET LORs are characterized by radial, azimuthal, copolar, and axial coordinates (r, φ, θ, z), the Orlov sphere is used to represent a “view” of this dataset, i.e. the collection of all (r, z) values in the specified angular direction (θ,φ). Any Orlov-sphere coordinate (φ,θ) in excess of φ=180 degrees, or π radians, represents the same information as the antipodal point (φ-π,−θ), since the two points represent the same line through space, except drawn in the opposite direction.
3D PET measurements are traditionally gathered into segments, represented by the segment curves in
Instead of using the coordinates (x_{s}, y_{s}, z_{s}) described above, the present invention uses other coordinates. According to one embodiment of the invention, a tilted plane can be used as shown in
As suggested by
Again, the Orlov sphere is useful to explain the data organization. The inventive concept is to span the same space of measurements (0 to 180 degrees in azimuth, and a fixed range of the copolar angle) with a set of curves, as shown by
Although the space we work with is three dimensional, this approach is two dimensional, in the sense that each curve corresponds to measurements across parallel planes as shown in
We will use the following definition for the tilted planes. Let (x, y, z) be coordinates of a point in the ordinary space used by PET scanners today. A tilted plane is defined by its z intercept, by a specified tilt angle τ, and by an orientation angle ψ. We will continue to use the same x and y coordinates for points in the tilted planes, but we identify the third spatial dimension through a parameter z′(ψ) which represents the z-intercept, in the original space, of the tilted plane:
z′(ψ)=z+tan τ×(y cos ψ−x sin ψ) (1)
The (x, y, z′(ψ)) coordinates are not orthogonal. Nevertheless, one can represent points in space uniquely in this representation, and fast linogram transformations can be performed with suitable Jacobian transformations to relate normal 3D Cartesian coordinates to the tilted-plane coordinates.
This tilted-planes approach to 3D PET has the following features.
(1) Organization of the Measured Projections
In the data recording procedure, or alternatively as a post-processing step, the LORs are organized so as to lie on planes identified, on the Orlov sphere, with tilted equators. This approach to organizing the data will work well in the case of continuous or quasi-continuous motion of the patient bed. Moreover, LORs can be directly transformed into a linogram representation.
(2) Modified EM-ADRT Reconstruction
The EM-ADRT algorithm, as described in the '065 published application cited above, is exercised, with modifications:
Specifically, to summarize from said '065 publication, a two-dimensional EM (2D EM) reconstruction algorithm is driven by a linogram L_{u,v }of size 3N×(2N−4), where N is a power of 2, for each tilt plane. Each pixel in the matrix of size N×N is given an initial value. Pixel values (x,y) are denoted in iteration i by the symbol λ_{x,y*} ^{i}.
A reciprocal back projection of the projection weights is then formed from the following equation:
where A^{T }is an adjoint ADRT projector, or back projector of the ADRT, and B_{x,y }is referred to as the normalization image. By modifying this equation, normalization and attenuation-weighting can be included in the reconstruction in order to achieve a better image quality.
Next, a loop is begun over iterations i. The estimate is then forward projected into linogram coordinates using the equation:
P_{u,v} ^{i}=A{λ_{x,y} ^{i}}
where A is the ADRT forward projector.
In all linogram bins (u,v) the correction ratio:
if P_{u,v} ^{i }exceeds the value ε, C_{u,v}=0 otherwise
is formed. In this equation, ε is a small constant value.
The correction factors are back projected and the normalization image is applied, as in the following equation:
λ_{x,y} ^{i+1}=λ_{x,y} ^{i}B_{x,y}A^{T}{C_{u,v}}
Finally, the loop described above is closed. Specifically, after the correction factors are back projected and the normalization image is applied, the method returns to the step of back projecting the projection weights.
Forward projection of a half-image I(x,y) at angles between 0 and π/4 radians is accomplished as follows. Two image buffers, R, with (N_{X}+N_{Y}) columns and N_{Y }rows are defined. The additional columns are needed because, although the linogram projection of rectangular space spans only N_{X }bins at zero degrees, an additional N_{Y }bins are needed for the projection at angles up to π/4 radians. One is identified as the “current” and one as the “previous” version of the buffer. Image values are loaded into the left columns of the “previous” buffer, leaving the rightmost N_{X }columns empty.
(3) OSEM-ADRT Reconstruction
An ordered-subsets version of the algorithm is created when we introduce an outer looping over groups of tilted-plane orientations. For example, with reference to
The simple z interpolation is an important feature of the new approach. Interpolation in the direction of just one axis is computationally simpler than 3D or 2D interpolation, normally required for fully 3D PET. These interpolations are a natural approach to removing distortions that arise when one uses the normal 3D PET organization (
While illustrative embodiments of the invention have been described herein, the present invention is not limited to the various preferred embodiments described herein, but includes any and all embodiments having equivalent elements, modifications, omissions, combinations (e.g., of aspects across various embodiments), adaptations and/or alterations as would be appreciated by those in the art based on the present disclosure. The limitations in the claims (e.g., including that to be later added) are to be interpreted broadly based on the language employed in the claims and not limited to examples described in the present specification or during the prosecution of the application, which examples are to be construed as non-exclusive. For example, in the present disclosure, the term “preferably” is non-exclusive and means “preferably, but not limited to.” In this disclosure and during the prosecution of this application, means-plus-function or step-plus-function limitations will only be employed where for a specific claim limitation all of the following conditions are present in that limitation: a) “means for” or “step for” is expressly recited; b) a corresponding function is expressly recited; and c) structure, material or acts that support that structure are not recited. In this disclosure and during the prosecution of this application, the terminology “present invention” or “invention” may be used as a reference to one or more aspect within the present disclosure. The language present invention or invention should not be improperly interpreted as an identification of criticality, should not be improperly interpreted as applying across all aspects or embodiments (i.e., it should be understood that the present invention has a number of aspects and embodiments), and should not be improperly interpreted as limiting the scope of the application or claims. In this disclosure and during the prosecution of this application, the terminology “embodiment” can be used to describe any aspect, feature, process or step, any combination thereof, and/or any portion thereof, etc. In some examples, various embodiments may include overlapping features. In this disclosure, the following abbreviated terminology may be employed: “e.g.” which means “for example.”
Cited Patent | Filing date | Publication date | Applicant | Title |
---|---|---|---|---|
US7120283 * | Jan 12, 2004 | Oct 10, 2006 | Mercury Computer Systems, Inc. | Methods and apparatus for back-projection and forward-projection |
US7394927 * | Apr 26, 2004 | Jul 1, 2008 | Koninklijke Philips Electronics N.V. | Iterative method of determining a spatial distribution of values of a property |
US20030190065 * | Mar 26, 2003 | Oct 9, 2003 | Cti Pet Systems, Inc. | Fast iterative image reconstruction from linograms |
Reference | ||
---|---|---|
1 | * | "3D-reconstruction using cone-beam backprojection, the Radon transform and linogram techniques", by Axelsson-Jacobson, C.; Defrise, M.; Danielsson, P.-E.; Clack, R.; Noo, F.; Nuclear Science Symposium and Medical Imaging Conference Record, 1995., 1995 IEEE vol. 3; Digital Object Identifier: 10.1109/NSSMIC.1995.500247; Publication Year: 1995 , P. |
2 | * | "Fast fully 3-D image reconstruction in PET using planograms", by Brasse, D.; Kinahan, P.E.; Clackdoyle, R.; Defrise, M.; Comtat, C.; Townsend, D.W.; Medical Imaging, IEEE Transactions on vol. 23 , Issue: 4; Digital Object Identifier: 10.1109/TMI.2004.824231; Publication Year: 2004 , pp. 413-425. |
3 | * | "Fast PET EM reconstruction from linograms", by Hamill, J.; Michel, C.; Kinahan, P.; Nuclear Science, IEEE Transactions on vol. 50 , Issue: 5 , Part: 2; Digital Object Identifier: 10.1109/TNS.2003.817405; Publication Year: 2003 , pp. 1630-1635. |
4 | * | "Linograms in Image Reconstruction from Projections", by Edholm, Paul R.; Herman, Gabor T.; Medical Imaging, IEEE Transactions on vol. 6 , Issue: 4; Digital Object Identifier: 10.1109/TMI.1987.4307847; Publication Year: 1987 , pp. 301-307. |
Citing Patent | Filing date | Publication date | Applicant | Title |
---|---|---|---|---|
WO2014016626A1 | Jul 23, 2012 | Jan 30, 2014 | Mediso Orvosi Berendezés Fejlesztö És Szerviz Kft. | Method, computer readable medium and system for tomographic reconstruction |
U.S. Classification | 382/131, 382/128 |
International Classification | G06K9/00 |
Cooperative Classification | G06T2211/424, G06T11/006 |
European Classification | G06T11/00T3 |
Date | Code | Event | Description |
---|---|---|---|
May 9, 2007 | AS | Assignment | Owner name: SIEMENS MEDICAL SOLUTIONS USA, INC., PENNSYLVANIA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:HAMILL, JAMES J.;REEL/FRAME:019269/0413 Effective date: 20070504 |
Mar 14, 2014 | REMI | Maintenance fee reminder mailed | |
Aug 3, 2014 | LAPS | Lapse for failure to pay maintenance fees | |
Sep 23, 2014 | FP | Expired due to failure to pay maintenance fee | Effective date: 20140803 |